Abstract:
Reversible Data hiding (RDH) is an evolving forensic and covert-communication technology that embeds data into a cover image (or other media like video or audio) so that
the embedded data later can be extracted for the copyright protection, integrity establishment or annotation. Developing such a scheme with better rate-distortion performance is challenging since a higher embedding rate usually causes more distortion in
the embedded image. Recently, the pixel-value-ordering (PVO) based RDH schemes
have shown better rate-distortion performance so far. However, the existing PVObased RDH schemes have not considered a suitable scanning order in a kernel that can
further improve the embedding rate-distortion performance.
This thesis, therefore, contributes to the development of a PVO-based RDH scheme
with a new PVO-kernel and backward embedding technique. Firstly, a new triangular kernel is proposed that captures the pixels correlated in the horizontal, vertical
and diagonal directions simultaneously. The proposed kernel is employed in a prominent PVO-based RDH scheme and is verified for a better (or occasionally similar)
rate-distortion performance than the existing schemes that rely on only the columnor row-kernel. Additionally, a new backward embedding technique is introduced to
counterbalance the distortion caused in a forward embedding phase.
Besides, the computational modeling, and the evaluation, analysis, and validation
of the new PVO-based RDH scheme are presented in the thesis. The simulation results
has demonstrated a promising performance of the proposed scheme and its improvement over the popular and state-of-the-art PVO-based RDH schemes. Particularly, a
significantly better rate-distortion performance is obtained at the higher embedding
rate, which means the proposed scheme is more promising to the applications with a
high embedding capacity requirement like electronic patient record hiding in medical
images. Moreover, the proposed RDH scheme developed using the new kernel and
backward-embedding would create a new paradigm of RDH for the future data hiding
research.
vii
